Abstract. We evaluate the GEOS-Chem atmospheric transport model (v8-02-01) of CO₂ over 2003–2006, driven by GEOS-4 and GEOS-5 meteorology from the NASA Goddard Global Modeling and Assimilation Office, using surface, aircraft and space-borne concentration measurements of CO₂. We use an established ensemble Kalman Filter to estimate a posteriori biospheric+biomass burning (BS + BB) and oceanic (OC) CO₂ fluxes from 22 geographical regions, following the TransCom-3 protocol, using boundary layer CO₂ data from a subset of GLOBALVIEW surface sites. Global annual net BS + BB + OC CO₂ fluxes over 2004–2006 for GEOS-4 (GEOS-5) meteorology are −4.4 ± 0.9 (−4.2 ± 0.9), −3.9 ± 0.9 (−4.5 ± 0.9), and −5.2 ± 0.9 (−4.9 ± 0.9) PgC yr⁻¹, respectively. After taking into account anthropogenic fossil fuel and bio-fuel emissions, the global annual net CO₂ emissions for 2004–2006 are estimated to be 4.0 ± 0.9 (4.2 ± 0.9), 4.8 ± 0.9 (4.2 ± 0.9), and 3.8 ± 0.9 (4.1 ± 0.9) PgC yr⁻¹, respectively. The estimated 3-yr total net emission for GEOS-4 (GEOS-5) meteorology is equal to 12.5 (12.4) PgC, agreeing with other recent top-down estimates (12–13 PgC). The regional a posteriori fluxes are broadly consistent in the sign and magnitude of the TransCom-3 study for 1992–1996, but we find larger net sinks over northern and southern continents. We find large departures from our a priori over Europe during summer 2003, over temperate Eurasia during 2004, and over North America during 2005, reflecting an incomplete description of terrestrial carbon dynamics. We find GEOS-4 (GEOS-5) a posteriori CO₂ concentrations reproduce the observed surface trend of 1.91–2.43 ppm yr⁻¹ (parts per million per year), depending on latitude, within 0.15 ppm yr⁻¹ (0.2 ppm yr⁻¹) and the seasonal cycle within 0.2 ppm (0.2 ppm) at all latitudes. We find the a posteriori model reproduces the aircraft vertical profile measurements of CO₂ over North America and Siberia generally within 1.5 ppm in the free and upper troposphere but can be biased by up to 4–5 ppm in the boundary layer at the start and end of the growing season. The model has a small negative bias in the free troposphere CO₂ trend (1.95–2.19 ppm yr⁻¹) compared to AIRS data which has a trend of 2.21–2.63 ppm yr⁻¹ during 2004–2006, consistent with surface data. Model CO₂ concentrations in the upper troposphere, evaluated using CONTRAIL (Comprehensive Observation Network for TRace gases by AIrLiner) aircraft measurements, reproduce the magnitude and phase of the seasonal cycle of CO₂ in both hemispheres. We generally find that the GEOS meteorology reproduces much of the observed tropospheric CO₂ variability, suggesting that these meteorological fields will help make significant progress in understanding carbon fluxes as more data become available.